Skip to content

[Performance] Compatible with flashmask API rename upgrade #9019

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Sep 30, 2024
Merged

[Performance] Compatible with flashmask API rename upgrade #9019

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 18 additions & 8 deletions paddlenlp/transformers/llama/fusion_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -214,16 +214,26 @@
)
else:
if attn_mask_startend_row_indices is not None:
assert alibi is None, "flash_attention_with_sparse_mask not support alibi"
assert alibi is None, "flashmask_attention or flash_attention_with_sparse_mask not support alibi"

Check warning on line 217 in paddlenlp/transformers/llama/fusion_ops.py

View check run for this annotation

Codecov / codecov/patch

paddlenlp/transformers/llama/fusion_ops.py#L217 

Added line #L217 was not covered by tests
if len(attn_mask_startend_row_indices.shape) == 2:
attn_mask_startend_row_indices = paddle.unsqueeze(attn_mask_startend_row_indices, axis=1)
attn_output = F.flash_attention_with_sparse_mask(
query_states,
key_states,
value_states,
attn_mask_start_row_indices=attn_mask_startend_row_indices,
is_causal=True,
)

if hasattr(F, "flashmask_attention"):
attn_output = F.flashmask_attention(

Check warning on line 222 in paddlenlp/transformers/llama/fusion_ops.py

View check run for this annotation

Codecov / codecov/patch

paddlenlp/transformers/llama/fusion_ops.py#L221-L222 

Added lines #L221 - L222 were not covered by tests
query_states,
key_states,
value_states,
startend_row_indices=attn_mask_startend_row_indices.unsqueeze(-1),
causal=True,
)
else:
attn_output = F.flash_attention_with_sparse_mask(

Check warning on line 230 in paddlenlp/transformers/llama/fusion_ops.py

View check run for this annotation

Codecov / codecov/patch

paddlenlp/transformers/llama/fusion_ops.py#L230 

Added line #L230 was not covered by tests
query_states,
key_states,
value_states,
attn_mask_start_row_indices=attn_mask_startend_row_indices,
is_causal=True,
)
else:
attn_output = F.scaled_dot_product_attention(
query_states,
Expand Down
Loading